Originally published at htpbe.tech. The version on htpbe.tech stays in sync with the latest detection algorithm — refer to it for the canonical text.

Document fraud detection is not one problem. It is at least three, and the tool that solves one of them is often useless for the others. If you are evaluating HTPBE?, iDenfy, or the manual review process your team has been running since the company was founded, the most important thing to understand before comparing prices is which problem each approach is designed to solve.

This article is a practical comparison of all three. It includes cost data, setup timelines, accuracy benchmarks, and a decision matrix you can use immediately. The goal is not to declare a winner — it is to help you identify which combination of tools is correct for your specific use case.

The Three Approaches at a Glance

Before getting into the details, here is the conceptual distinction that matters most:

• Manual review answers: "Does a human expert believe this document is legitimate?"
• KYC/identity proofing platforms (iDenfy, Onfido, IDWise, Jumio, Ondato) answer: "Does this document look like a valid identity document, and does this person match it?"
• HTPBE? answers: "Was this specific PDF file modified after it was created?"

These are related questions, but they are not the same question. The first requires human judgment. The second requires template and biometric analysis. The third requires metadata inspection. None of the three tools is interchangeable with the others — and understanding this clearly is what makes the comparison actionable.

Approach 1: Manual Review

What It Is

Manual document review means a trained analyst — often a compliance officer, a document examiner, or a fraud investigator — reads a submitted document and makes a judgment about its authenticity. In high-stakes environments, this may involve a certified forensic document examiner using physical or digital forensic tools and producing a written opinion.

Cost

Manual review costs $50–$150 per document in analyst time at a loaded rate of $30–$75 per hour, assuming 10–20 minutes per document. This estimate holds for routine compliance review. Forensic examinations that generate expert witness testimony for legal proceedings cost substantially more — typically $200–$500 per document and up.

Speed

A document routed to a review queue can take anywhere from 10 minutes to several days, depending on queue volume, analyst availability, and document complexity. There is no throughput ceiling per analyst, and throughput does not scale without hiring.

Accuracy

Human reviewers catch roughly 60–70% of fraudulent documents on average, according to studies on document fraud detection. Experienced forensic examiners perform meaningfully better, but their involvement is reserved for high-value or legally significant cases and cannot be applied at volume.

What It Does Well

Manual review is the correct choice in two narrow but important scenarios:

1. Legally significant documents requiring expert witness testimony. If the outcome of a dispute depends on a document’s authenticity and the matter will be argued in court, only a certified forensic examiner produces output that is admissible and defensible.
2. Documents requiring human chain of custody. Some regulatory or contractual frameworks require that a named individual attest to having reviewed a document. No automated system can substitute for this.

What It Does Poorly

Manual review cannot handle volume. It does not produce a machine-readable audit trail unless someone creates one manually. It is inconsistent across reviewers, time of day, and individual fatigue. At $100 per document and 200 documents per month, you are spending $20,000 monthly on review before any other costs.

For anything requiring speed, volume, or cost efficiency, manual review is a last resort, not a workflow.

Approach 2: KYC and Identity Proofing Platforms

What They Are

The major identity proofing providers — iDenfy, IDWise, Onfido, Jumio, Ondato — offer what is commonly called KYC (Know Your Customer) document fraud detection. These platforms are genuinely sophisticated, well-engineered products that solve a specific and important problem.

What They Actually Check

KYC document fraud detection typically combines several checks:

• Document template validation — Does this ID conform to the layout, typography, security features, and design patterns of a legitimate government-issued document from this jurisdiction?
• Liveness detection — Is the person submitting this a live human being, not a photograph or video replay?
• Face match — Does the face captured during the liveness check match the face on the submitted identity document?
• Field extraction and cross-referencing — Do the extracted name, date of birth, and document number match other data in the application?

What KYC platforms are answering, in every case, is a version of this question: "Is this a real identity document, and does it belong to the person submitting it?"

What They Do Not Check

KYC platforms are not designed to answer a different question: "Was this specific PDF file modified after it was created?"

Consider a common fraud pattern in lending. An applicant downloads their real bank statement from their bank portal — the template is 100% legitimate, the account number is real, the bank logo and formatting are correct. They open the PDF in Microsoft Excel, change the account balance from $4,200 to $42,000, and export it back as a PDF. They upload it to your application portal.

A KYC platform inspects the document visually: template checks out, formatting is correct, the name matches — cleared. The file has now passed KYC validation despite having been produced by Excel rather than the bank’s document generation system. The metadata says so explicitly; the KYC platform simply is not reading it.

This is not a flaw in iDenfy or Onfido. It is a scope mismatch. These platforms are built to check identity documents — passports, national IDs, driver’s licenses. They are not built to inspect arbitrary PDFs submitted as financial supporting documents.

Cost and Setup

KYC platform pricing typically runs $0.50–$1.50 per fraud detection, with volume discounts at higher tiers. Setup requires enterprise onboarding: a sales cycle of 4–12 weeks, legal agreements, and technical integration with your document submission flow. Minimum contract commitments of $500–$1,000 per month are common, particularly for platforms targeting regulated financial services.

Best For

• User onboarding flows requiring identity proofing
• KYC/AML regulatory compliance for financial services
• Any workflow where the core question is "is this person who they claim to be?"

Not Best For

• Checking whether a specific PDF received via email has been tampered with
• Pre-screening financial supporting documents (bank statements, pay stubs, invoices) for file-level modification
• Any workflow where you already know who the person is and only need to check a document they submitted

Approach 3: HTPBE?

What It Is

HTPBE? is a PDF integrity fraud detection API. It analyzes the metadata of a submitted PDF — creation timestamps, producer software, creator application, incremental update chains, cross-reference table structure — and returns a verdict indicating whether the file was modified after it was originally created, and whether its origin is consistent with its claimed provenance.

What It Returns

Every analysis returns one of three verdicts:

• intact — No post-creation modifications detected; origin metadata is consistent with the document type
• modified — Post-creation modifications are detected via incremental update chains, timestamp anomalies, or tool signature mismatches
• inconclusive — The file cannot be checked as institutionally generated; consumer software origin detected

The inconclusive verdict is operationally significant. A bank statement produced by Microsoft Excel is inconclusive because HTPBE? cannot prove that specific numbers were changed — but it can prove the file was produced by Excel, not by banking software. For a bank statement, these two facts are mutually exclusive. The signal is actionable.

Cost and Setup

HTPBE? pricing runs $0.33–$0.50 per check depending on plan:

• Starter: $15/month for 30 checks ($0.50/check)
• Growth: $149/month for 350 checks ($0.43/check)
• Pro: $499/month for 1,500 checks ($0.33/check)
• Enterprise: custom pricing with on-premise deployment (Docker/Kubernetes)

Setup is a single API call. There is no enterprise onboarding, no minimum contract, no sales cycle. A Growth plan API key is provisioned at signup. A typical integration — getting a document URL, posting it to the HTPBE endpoint, and handling the JSON response — takes under 30 minutes.

curl -X POST https://api.htpbe.tech/v1/analyze \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://your-storage.com/document.pdf"}'

What It Does Well

HTPBE? is the correct tool when the question is: "Was this specific PDF file modified after it was created?" Use cases where this question matters:

• Invoice fraud detection — Before authorizing payment on a vendor invoice, check the PDF has not been modified since issuance
• Contract integrity — Altered contract PDFs change payment terms or liability clauses after signing; catching this before filing prevents costly disputes
• Insurance claims documents — Claims fraud via altered PDFs inflates repair estimates, medical bills, and receipts before submission
• Bank statement pre-screening — Flag documents produced by consumer software before they reach underwriting
• Diploma and certificate integrity — Pre-screen incoming credentials before investing in official fraud detection
• Bulk document ingestion — Run every incoming PDF through a first filter at $0.43 before routing to more expensive processes

What It Does Not Do

HTPBE? does not perform face matching, liveness detection, identity document template validation, OCR of document content, or any form of identity proofing. If the question is "is this person who they say they are?", HTPBE? is the wrong tool. You need a KYC platform for that.

The "When to Use What" Matrix

The Complementary Stack: HTPBE? + KYC

For fintech and lending teams, the financially optimal configuration is not "HTPBE? instead of KYC" — it is HTPBE? as a cheap first filter before KYC.

The logic is straightforward. Not every document submitted to a loan application needs full KYC treatment. Many are clearly fraudulent at the file level, long before any identity matching is attempted. Running those documents through a $1.50/fraud detection KYC check is unnecessary cost.

At 500 documents per month:

The hybrid model provides equivalent protection — or better, since it closes the file-level modification gap that KYC alone misses — at 40–70% lower cost. The savings compound at higher volumes.

TCO Calculation: 200 Documents Per Month Over 12 Months

These numbers are for 200 documents per month, a realistic volume for a mid-size lending operation, HR team, or property management company. The manual review line is included for completeness; it is almost never the right answer at this volume except for specific legally-mandated document types.

Note that the KYC annual cost excludes the platform minimum contract and onboarding costs, which can add $6,000–$12,000 in year one. HTPBE? has no minimum contract and no onboarding cost.

Decision Criteria

Use this to route your specific situation:

“Was this specific PDF modified?”
HTPBE?. This is its precise design purpose. The verdict, modification confidence level, producer metadata, and timestamp analysis are all oriented around this question.

“Is this the right person submitting this document?”
KYC platform. Face match, liveness detection, and identity document template validation are the tools for this. HTPBE? does not touch identity.

“I need to testify about document integrity in court.”
Manual forensics expert. Certified forensic document examiners produce admissible written opinions. HTPBE? produces a machine-readable API response, which is useful evidence but is not a substitution for a human expert’s sworn testimony.

“GDPR or HIPAA compliance requires on-premise processing.”
HTPBE? Enterprise. The enterprise tier offers Docker and Kubernetes deployment so the analysis runs entirely within your infrastructure. No document data leaves your environment.

“Less than 5 documents per day, no API integration required.”
HTPBE? web tool. Free to try — new accounts get 5 free checks. Upload a PDF and sign in to receive the analysis; after that it is pay-per-check (credit packs from $5) or a subscription from $15/mo. Suitable for ad-hoc fraud detection without an engineering investment.

“500+ documents per day.”
HTPBE? Pro ($499/month for 1,500 checks) or Enterprise. At this volume, the per-check savings at Pro ($0.33 vs. $0.43 at Growth) begin to matter, and Enterprise pricing can be negotiated for significant additional savings.

“I need both identity proofing and file integrity checking.”
HTPBE? + KYC platform in sequence. Run HTPBE? first as a cheap pre-filter. Escalate documents that pass (or that return inconclusive) to KYC for identity proofing. Documents that return modified from HTPBE? are rejected before incurring KYC cost.

A Note on Accuracy and Reliability

Manual review accuracy varies significantly by reviewer experience and document type: roughly 60–70% detection rates for trained reviewers on average, higher for specialists on specific document types they examine frequently.

KYC platforms have published accuracy rates in the 95%+ range for identity document template validation on major ID types. These numbers hold for the task they are designed for — validating that a document looks legitimate. They do not apply to the file-level modification detection task.

HTPBE? operates on a different axis. Metadata analysis is deterministic: the Producer field either says "Microsoft Excel" or it does not. Timestamp anomalies either exist or they do not. The modification_confidence field in the response reflects the strength of the structural evidence — "certain" means multiple independent signals converge on the same conclusion. A modification_confidence: "certain" on a modified verdict means the structural evidence for modification is unambiguous, not a probabilistic estimate.

This distinction matters for how you integrate the output. HTPBE? verdicts are most useful as routing signals, not as binary pass/fail gates. A modified response routes a document to review queue. An intact response allows it to proceed. An inconclusive response triggers an alternative sourcing request for the document.

Summary

The question “which document fraud detection tool should we use?” almost always has the answer: “which problem are you trying to solve?”

For identity proofing and regulatory KYC compliance, iDenfy, Onfido, and equivalent KYC platforms are the right tools. They are built for this and they are good at it.

For expert testimony on document integrity in legal proceedings, a certified forensic document examiner is the only option. No API produces admissible courtroom evidence.

For checking that specific PDF files have not been modified after creation — invoices, bank statements, diplomas, pay stubs, contracts — HTPBE? is the purpose-built tool. At $0.43 per check, with 30-minute setup and no minimum contract, it is also the lowest-cost entry point in the document fraud detection space.

For most teams handling financial documents at volume, the optimal configuration is HTPBE? as a first filter plus a KYC platform for identity proofing when identity matters. This closes both the file-integrity gap and the identity gap, at lower total cost than KYC alone.